Representative analysts, supervisors, controllers, FB&T leaders, and technology operators
Enterprise Runtime Topology
bp Sphere is the finance runtime layer above systems of record and systems of execution. It operates across identity, security, context, policies, agents, evidence, replay, model routing, resilience, and enterprise systems without replacing SAP, Ariba, FSCM, BlackLine, Workiva, ServiceNow, MDG, Bottomline, UDP, Power BI, ECC, or S/4HANA.
Finance, control, evidence, replay, forecast, credit, treasury, and platform agents
Representative adapters now; BP production connectors during implementation
Authority, SOX, cyber, approval, data-access, and AI-governance controls
Generated, assisted, or governed decision-record shape
Representative topology with bounded degraded-mode design
Reference deterministic-retrieval and caching design metric
Reference target: RTO 10 min · RPO 5 min
Human accountability layer for analysts, supervisors, controllers, FB&T leaders, CFO, and administrators.
Authentication, authorization, role resolution, agent identity, token lifecycle, and secrets access.
Runtime enforcement for authority, SOX, segregation of duties, compliance, AI governance, and write boundaries.
Receives enterprise events and turns them into normalized business signals that activate missions and agents.
Resolves enterprise meaning through ontology, graph relationships, glossary, vector retrieval, and history.
Builds trust through evidence packs, source traceability, document links, transaction lineage, and replay records.
Runs specialized agents through policy-gated tools, bounded autonomy levels, cost controls, and replay.
Canonical business objects shared by agents, evidence, policies, decisions, actions, and replay.
Adapter contracts emulate enterprise integrations in validation and connect to BP systems in production.
Systems of record remain the source of truth; bp Sphere operates above them as a decision and action runtime.
The BP atlases are explicit: remove avoidable work first, standardize the remaining variants, automate clean repeatable work, and govern material actions with evidence, policy, HITL, replay, and value attribution.
Current runtime proof includes governed chat, WIIF assistant portability, token economy, fail-closed policy enforcement, deterministic replay, signed evidence, runtime seam adoption audit, learning promotion, autonomy demotion enforcement, shared case-state propagation, and observability readiness.
| Capability | Status | Runtime proof | Surface / endpoint |
|---|---|---|---|
| Experience Orchestration Layer | Surface available | Shared decision objects render across desktop, mobile, and chat surfaces. | /ui/mission/p2p, /ui/mobile-cockpit, /ui/enterprise-chat |
| Governed Chat Interface | Surface available | Chat resolves role, mission, case, evidence, policy, action contract, replay, and feedback. | /ui/enterprise-chat |
| Workforce Intelligence Integration Fabric | Surface available | Assistant registry, connector packages, certification, compliance mappings, identity resolution, failover drills, and workforce coordination APIs. | /ui/workforce-intelligence-fabric, /ui/assistant-operations-center |
| Token Economy Control Plane | Surface available | LLM calls are routed, cached, denied, downgraded, or escalated before spend occurs; token value is attributed to decisions. | /ui/ai-cost-usage, /api/iaf/decision-intelligence/token-economy/decision |
| Guardrail Policy Engine | Surface available | Policy evaluation fails closed in production posture; non-fail-closed failures route to escalation instead of silent execution. | /ui/policy-registry, /ui/control-plane |
| Decision Replay + LLM Determinism | Surface available | Execution seed, prompt/context hash, evidence hash, policy version, and replay records are persisted and validated. | /ui/decision-journey, /ui/single-decision-audit, /ui/decision-replay-studio |
| Externally Verifiable Evidence | Surface available | Evidence bundles are signed with asymmetric verification and can be checked through the verify endpoint. | /ui/evidence, /api/iaf/decision-intelligence/evidence/verify |
| Runtime Adoption Audit | Endpoint available | Agents/workflows are audited for identity, policy, evidence, replay, value, token, and learning seam adoption. | /api/iaf/decision-intelligence/runtime-adoption-audit |
| Skill-Binding Learning Promotion | Endpoint available | Skill learning overlays require evidence, confidence, and runtime effect before promotion into runtime metadata. | /api/iaf/decision-intelligence/skill-learning-overlays/{id}/validate |
| Autonomy Demotion Enforcement | Endpoint available | Telemetry demotion guidance writes supervisory state, event logs, and owner review queue items. | /api/iaf/decision-intelligence/autonomy/demotions/enforce |
| Shared Case-State Propagation | Endpoint available | One case resolution updates analyst state, supervisor state, value records, replay reference, and CFO rollup readiness. | /api/iaf/decision-intelligence/case-state/{case_id}/rollup |
| Observability Readiness | Endpoint available, env-bound | Production telemetry posture exposes OTEL endpoint status, Prometheus/failure-mode surfaces, and SLO targets. | /api/iaf/decision-intelligence/observability/readiness |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| SAP Ariba PR / PO creation and supplier workflow | Strategic | |
| SAP ECC Invoice posting across 12+ active instances | Live | HWG, WRC, PRC, WRO, PRO, PRI, WRI, PRL, PRE, PRV, PR4, WR4 |
| JDE Castrol / C&P ERP outside near-term S/4 migration | Live | |
| SAP S/4HANA Quantum future ERP target | Future | Castellon and Kwinana Q3 2026 wave |
| SAP VIM / OpenText Current invoice processing | Retiring | |
| Appian Project Elevate Invoice automation extension | In-flight | MVP with 98 FTE benefit identified |
| ServiceNow Exception routing and ticketing | In-flight | |
| SAP GRC / DofA SoD and authority matrix | Live | |
| SAP MDG + Semarchy Vendor master governance | Strategic | |
| Bottomline TMS Payment and treasury execution context | Strategic | |
| UDP / Databricks / Power BI Analytics and control tower | Strategic |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| SAP FSCM Credit | Strategic | |
| SAP ECC Billing and AR subledger across multiple instances | Live | |
| BlackLine Auto Cash Matching Cash matching PoC and future automation seam | PoC | |
| Aravo Customer KYC / CDD and discovery | In-flight | |
| Appian / IBOAT O2C automation and workflow | Live | |
| Process HQ Process intelligence | Live | |
| SAP MDG + Semarchy Customer master remediation | Strategic | |
| ServiceNow / Salesforce Chatbot Customer query and case surface | In-flight | |
| SAP GRC + ServiceNow IRM Controls | Live | |
| UDP / Databricks / Power BI AR analytics and tower reporting | Strategic |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| SAP ECC Current journal / close source | Live | ~30k manual journals per quarter |
| SAP CFIN Central Finance harmonized truth | Live Jan 2026 | |
| SAP Group Reporting Group reporting future state | Live Apr 2026 | |
| SAP FCC / FBW Legacy close and reporting tools | Retiring | |
| Workiva Statutory and reporting evidence | In-flight | |
| BlackLine / LIBRA Reconciliation and intercompany | Strategic | |
| Nakisa Lease accounting | Live | $14bn lease portfolio |
| SAP BTP / PaPM Auto-capitalization and allocations | Rolling out | |
| SAP GRC + ServiceNow IRM Controls | Live |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| SAP IBP Future planning | In-flight | |
| SAP BPC Legacy planning and budgeting | Retiring | |
| SAC Analytics and spend cube PoC | PoC | |
| CFIN / Group Reporting Actuals source | Live | |
| SAP BW + Power BI Current reporting | Transitioning to UDP |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| Bottomline TMS Treasury management | Strategic | |
| C&B Hub Cash and banking | Complete | |
| ISO 20022 / CAMT.054 Banking and remittance standard | 90% deployed | |
| SAP S/4 Treasury Future treasury modules | Future | |
| SAP FSCM Cash management | Live |
Sphere operates above these systems as the decision, evidence, policy, replay, and value layer. Systems of record remain the source of truth.
| System | Status | Transition note |
|---|---|---|
| SAP MDG SAP master data | Strategic | |
| Semarchy Non-SAP master data | Strategic | |
| SAP GRC Access and SoD | Live | |
| ServiceNow IRM Controls tracking | Live | |
| Aravo KYC / sanctions / CDD | In-flight |
The P2P and O2C atlases show that policy simplification and data-quality gates must come before autonomy. These simulations are the first executable candidates for the Policy Workbench.
Widen PO tolerance from ±1% to ±3%
Current: ±1% tolerance creates avoidable normal price-variation exceptions.
Proposed: ±3% for Tier 1 and Tier 2 vendors covering roughly 70% of vendor base.
Expected impact: Eliminates roughly 15% of invoice exceptions with SAP ECC config and policy approval.
Auto-GR for repeat contracted services
Current: Forgotten GR creates invoice failure and GRIR backlog.
Proposed: Auto-post GR at service-period end unless dispute raised within 48h.
Expected impact: Attacks a major source of GRIR backlog at source.
Rationalize 40+ payment terms to five tiers
Current: Fragmented terms reduce DPO optimization and routing consistency.
Proposed: Net 30, Net 45, Net 60, Early Pay Discount, Immediate.
Expected impact: Supports working-capital optimization and simpler policy execution.
Structured remittance before cash lands
Current: No remittance drives unidentified cash and manual matching.
Proposed: Request remittance 48h prior and mandate ISO 20022 / CAMT.054 for top customers.
Expected impact: Eliminates roughly 70% of unidentified cash at source.
BPRR credit tier policy
Current: Periodic credit review consumes effort even for low-risk customers.
Proposed: BPRR 1-2 auto-renew, BPRR 3-4 annual review, BPRR 5+ event-triggered.
Expected impact: Removes a large share of low-risk credit review workload.
Standardize deductions to 15 global categories
Current: Inconsistent reason codes cause manual investigation and ownership confusion.
Proposed: Finance, Sales, and Commercial ownership by category.
Expected impact: Reduces manual deduction investigation by roughly 60%.
Value attribution should roll up transaction -> case -> agent -> process -> tower -> location -> FB&T -> CFO.
| Effort Release | 45-55% future-state reduction ambition. |
| Leakage Stopped | $139M annual duplicate exposure prevented at source. |
| Cash Unlocked | $200-300M working-capital opportunity plus DPO uplift. |
| Risk Reduced | 50%+ risk reduction ambition through evidence, controls, and replay. |
| Effort Release | 41% manual cash application targeted for prevention, standardization, and automation. |
| Leakage Stopped | Deductions and write-offs moved to evidence-backed owner workflows. |
| Cash Unlocked | Unapplied cash reduction from $116M average and unidentified cash from $83M peak. |
| Risk Reduced | BPRR, IFRS 15, write-off, and credit override boundaries remain human-governed. |
Click an agent to inspect identity, owner, version, permissions, policies, tools, connected systems, last decision, cost, and replay history.
| Agent | Category | Status | Version | Owner | Risk | Autonomy | Last execution | Cost |
|---|---|---|---|---|---|---|---|---|
| Duplicate Invoice Agent | P2P | Active | v3.4 | P2P Control Owner | High | Draft action | 14:22 | $0.018 |
| Journal Risk Agent | R2R | Active | v2.9 | Controller | High | Recommend | 14:19 | $0.026 |
| Credit Exposure Agent | O2C | Active | v2.7 | Credit Director | High | Human-approved | 14:18 | $0.031 |
| Forecast Driver Agent | FP&A | Active | v4.1 | FP&A Lead | Medium | Recommend | 14:21 | $0.024 |
| Liquidity Intelligence Agent | Treasury | Active | v2.3 | Treasurer | High | Draft action | 14:22 | $0.022 |
| Control Assurance Agent | Control | Active | v3.0 | SOX Owner | High | Block/escalate | 14:23 | $0.017 |
| Replay Integrity Agent | Platform | Active | v1.8 | Platform Owner | Medium | Read-only | 14:24 | $0.006 |
Click any object to show source lineage, ownership, freshness, evidence links, and replay relationships.
Click an adapter. Every adapter declares purpose, production target, authentication type, access mode, sync, latency, and the validation caveat.
Finance and procurement transactions
Universal journal and operational ERP
Central Finance harmonized truth
PO, supplier, contract, requisition signals
Reconciliations, close tasks, certifications
Curated data products, features, lineage
Incidents, work orders, change records
Documents, policies, approvals
Approval messages, supplier correspondence, exception evidence
Approved tool execution boundary
| Cost driver | Current view | Control mechanism |
|---|---|---|
| Token Consumption | 57B/month modeled | Frontier reasoning only when judgment is required |
| Model Cost | $200K-$400K/month scenario | Governed by mission budgets, routing, cache, compression |
| API Cost | Bounded | Adapters batch, cache, and use event triggers |
| SAP Calls | 6% verification calls | Read-first; write-back only after approval |
| Ariba Calls | 2% writeback calls | Workflow actions rather than uncontrolled mutation |
| Databricks Reads | 92% context retrieval | Curated data products for historical context |
| Cache Hits | 71% | Evidence, policy, and context reuse |
| Deterministic Decisions | 68% | Policy, retrieval, replay, and simulation run without LLM calls |
| Signal | Value | Control proof |
|---|---|---|
| Authentication Success | 99.8% | Local Entra-compatible validation; BP Entra/SSO production target |
| Authorization Failures | 31 | Denied actions retained in audit stream |
| Prompt Injection Attempts | 7 | Gateway blocks tool/prompt mutation attempts |
| Tool Injection Attempts | 2 | MCP/tool gateway denied unapproved tool path |
| Sensitive Data Access | 186 | Classified and policy-logged |
| Security Events | 12 | Forwardable to BP SIEM/SOC target |
| Dimension | Status | Design proof |
|---|---|---|
| Primary Runtime | Healthy | Prod namespace serving internal service |
| Secondary Runtime | Standby | Warm standby design; validation representation |
| Last Backup | 14:00 UTC | Postgres snapshots + evidence store manifests |
| Last Recovery Test | Yesterday | Replay manifest restored and verified |
| Recovery Readiness | Ready | Critical config, DB, evidence, and replay paths accounted for |
| Current RTO | 10 min | Target design for pilot profile |
| Current RPO | 5 min | Target design using event/replay snapshot cadence |
| Decision type | Preferred | Fallback | Local / sovereign fallback | Why selected |
|---|---|---|---|---|
| Judgment narrative | Claude Sonnet | GPT fallback | Local Llama/Qwen fallback | High-context explanation needs frontier model |
| Policy check | Deterministic | No LLM | No LLM | Rules engine is cheaper and more auditable |
| Evidence extraction | Small model / OCR | GPT mini | Local Mistral/Phi | Document extraction does not need premium reasoning |
| Scenario simulation | Deterministic + Python model | No LLM | No LLM | Finance model should be repeatable |
| Executive summary | Premium model | GPT fallback | Local Llama/Qwen/Mistral summary | Narrative quality matters, but output remains evidence-bound |
Search dimensions: decision, object, agent, policy, user, date. This example shows the trace BP architects should expect for every material decision.
| Replay step | Record | What it proves |
|---|---|---|
| Signal | REF-TICK-30577 | Invoice duplicate risk detected from SAP + Ariba representative adapters |
| Context | CTX-P2P-884 | Vendor, PO, invoice, payment run, prior duplicates, policy domain resolved |
| Policy | FIN-DOA-2026-07 | Authority and duplicate payment control evaluated |
| Evidence | EP-CTL-30577 | Invoice, PO, vendor master, payment run, approval matrix attached |
| Agent | Duplicate Invoice Agent | Match result, confidence, explanation, and action proposal stored |
| Decision | DEC-P2P-884 | Block payment and request supplier validation |
| Action | ACT-HOLD-771 | Draft payment hold prepared; human approval required |
| Outcome | Pending | Learning retained only after human outcome is known |